A common base four-finger InOaAs/InP double heterojunction bipolar transistor with 535 OHz fmax by using the 0.5 μm emitter technology is fabricated. Multi-finger design is used to increase the input current. Common ...A common base four-finger InOaAs/InP double heterojunction bipolar transistor with 535 OHz fmax by using the 0.5 μm emitter technology is fabricated. Multi-finger design is used to increase the input current. Common base configuration is compared with common emitter configuration, and shows a smaller K factor at high frequency span, indicating a larger breakpoint frequency of maximum stable gain/maximum available gain （MSG/MAG） and thus a higher gain near the cut-off frequency, which is useful in THz amplifier design.展开更多
In this paper, the efficiencies of two production lines, one is with commonly, the other is with separately buffer capacities are compared, the result is that the former one is better than the latter
基于强化学习的最大公共诱导子图(Maximum Common Induced Subgraph,MCIS)算法在处理历史搜索中低频出现的顶点时存在局限,难以评估其真实重要性并进行有效探索。为此,提出一种基于动作与环境反馈的前向预测方法。动作反馈通过奖励机制...基于强化学习的最大公共诱导子图(Maximum Common Induced Subgraph,MCIS)算法在处理历史搜索中低频出现的顶点时存在局限,难以评估其真实重要性并进行有效探索。为此,提出一种基于动作与环境反馈的前向预测方法。动作反馈通过奖励机制量化分支顶点的剪枝效果,环境反馈则用双域个数来表征待搜索子图的大小。前向预测通过单边采样选择顶点模拟分支,并根据反馈确定最佳顶点。实验结果表明,新算法McSplitLA比McSplitDAL多解决7个算例,平均求解时间减少11.2%~17.9%,有效提高了剪枝率并优化了探索方向。展开更多
基金Supported by the National Basic Research Program of China under Grant No 2011CB301900the Natural Science Foundation of Jiangsu Province under Grant Nos BK2011010 and BY2013077
文摘A common base four-finger InOaAs/InP double heterojunction bipolar transistor with 535 OHz fmax by using the 0.5 μm emitter technology is fabricated. Multi-finger design is used to increase the input current. Common base configuration is compared with common emitter configuration, and shows a smaller K factor at high frequency span, indicating a larger breakpoint frequency of maximum stable gain/maximum available gain （MSG/MAG） and thus a higher gain near the cut-off frequency, which is useful in THz amplifier design.
文摘In this paper, the efficiencies of two production lines, one is with commonly, the other is with separately buffer capacities are compared, the result is that the former one is better than the latter
文摘基于强化学习的最大公共诱导子图(Maximum Common Induced Subgraph,MCIS)算法在处理历史搜索中低频出现的顶点时存在局限,难以评估其真实重要性并进行有效探索。为此,提出一种基于动作与环境反馈的前向预测方法。动作反馈通过奖励机制量化分支顶点的剪枝效果,环境反馈则用双域个数来表征待搜索子图的大小。前向预测通过单边采样选择顶点模拟分支,并根据反馈确定最佳顶点。实验结果表明,新算法McSplitLA比McSplitDAL多解决7个算例,平均求解时间减少11.2%~17.9%,有效提高了剪枝率并优化了探索方向。
